Black has the kernel of a discovery on the d-file: his bishop masks his queen, which otherwise could take White’s queen. White’s queen is defended by its king, however, so QxQ would not yet be profitable. And there is another problem: What unmasking move could Black play? We look first for a check the bishop can inflict and find Bb4, but one always must consider what the reply to such a move would be; sometimes your opponent will have an answer that fends off the check and the unmasked threat. This is such a case: Bd2 would stop both of Black’s threats. (Another possible sequence would be 1. …Bb4+; 2. a3xB, QxQ; 3. KxQ, Nxf2+ (a knight fork); 4. Ke1, Nxh1; 5. Bg2 (where the knight has no escape)—and White is okay, as he has won a queen and two pieces in trade for a queen, and pawn, and a rook.)
Still, for Black to have the kernel of a discovery in place like this, with the enemy queen at one end of it, is an opportunity not to be abandoned lightly. Black looks for ways to move the White king by checking it with his other pieces and finds none. But there is another way to force a king to move: capture something on a square next to it—a square only the king protects, so that it has to move to recapture. Nxf2 comes to mind, as it not only captures a pawn but forks White’s queen and rook. Suppose White then plays KxN. Now how would the board look? The bishop on d6 would have a check with Bxg3+, and this time there would be no way for White to defend against it while also protecting the queen—which would have been left loose when the king moved away. So now the discovery works.